Reduced leaf complexity in tomato wiry mutants suggests a role for PHAN and KNOX genes in generating compound leaves

In higher plants, the shoot axis is radially symmetrical while lateral organs such as leaves have asymmetric features. This asymmetry is visible along three axes of the organ; proximodistal (base to tip), mediolateral(midrib to margin) and adaxial-abaxial (upper and lower). In most higher plants the upper (adaxial) part of the leaf is anatomically and physiologically different from the bottom (abaxial) part of the leaf. This asymmetry (ab-adaxiality) is established early in the leaf primordium and the shoot apical meristem (SAM) seems to provide positional cues for the initial establishment of this asymmetry (Hanawa,1961; Lynn et al.,1999; Snow and Snow,1959; Sussex,1954; Sussex,1955; Timmermans et al.,1998). Ab-adaxiality defective mutants have been reported in Antirrhinum (Waites and Hudson,1995), Arabidopsis(Bohmert et al., 1998; Bowman and Smyth, 1999; Chen et al., 1999; Eshed et al., 2001; McConnell and Barton, 1998; Sawa et al., 1999b; Siegfried et al., 1999), maize(Freeling, 1992; Timmermans et al., 1998), Nicotiana (McHale,1993a; McHale,1993b; McHale and Marcotrigiano, 1998), tomato(Kessler et al., 2001) and pea(Meicenheimer et al.,1983).

Several genes in several species are thought to specify the adaxial and abaxial domains. For example, leaf adaxial cell fate is replaced by abaxial cell fate in the phantastica mutation of Antirrhinum,suggesting that PHANTASTICA (PHAN), a MYB domain transcription factor, plays an important role in establishing (or maintaining)adaxial cell fate in leaf primordia(Waites and Hudson, 1995; Waites et al., 1998). In Arabidopsis ARGONAUTE1 (AGO), REVOLUTA (REV) and PINHEAD (PNH) are also important for specifying adaxial cell fate in lateral organs and for promoting meristematic activity in the SAM and axillary meristems (Bohmert et al.,1998; Lynn et al.,1999; Talbert et al.,1995). PHABULOSA and PHAVOLUTA are homeodomain-leucine zipper (HD ZIP III) proteins with a START (steroid/lipid-binding) domain expressed in the adaxial cells of the leaf primordium and in the SAM and semi-dominant mutations in these genes produce radial leaves with adaxial cell fates(McConnell and Barton, 1998; McConnell et al., 2001). In the leafbladeless mutant in maize, ectopic patches of abaxial identity are seen on the adaxial side of the leaf and ectopic lamina forms at the boundary between the two cell fates(Timmermans et al., 1998). FILAMENTOUS FLOWER (FIL), YABBY2 (YAB2), YABBY3 (YAB3) and KANADI are expressed only abaxially in all lateral organs of Arabidopsis, and ectopic expression of FIL or YAB3 is sufficient to induce ectopic abaxial patches in the adaxial region of the leaf(Sawa et al., 1999a; Siegfried et al., 1999). Together, all these mutant phenotypes strongly suggest that the juxtaposition of adaxial and abaxial cell fates is necessary for proper leaf lamina development in simple-leafed species, and that adaxial and abaxial cell fates are mutually antagonistic.

The Class I KNOTTED-1 LIKE HOMEOBOX (KNOX I) genes play an important role in maintaining indeterminacy in the SAM and in subsequent shoot development. Loss-of-function mutations in some of these genes (e.g. kn1 and stm) result in an inability to form or maintain a SAM (Barton and Poethig, 1993; Kerstetter et al., 1997; Smith et al., 1995; Vollbrecht et al., 2000). Mutations in other KNOX genes cause reduced internode or axis elongation (Postma-Haarsma et al.,2002; Venglat et al.,2002). Ectopic overexpression of KNOX genes in dicots leads to more dissected and highly lobed leaves, often accompanied by ectopic shoot meristem formation on leaves (Chen et al., 1997; Chuck et al.,1996; Janssen et al.,1998; Lincoln et al.,1994; Nishimura et al.,2000; Sinha et al.,1993). Dominant mutants in the KNOX gene LeT6, Mouse Ears (Me) and Curl (Cu), express LeT6ectopically in the mature leaves and show increased leaf dissection(Chen et al., 1997; Parnis et al., 1997). Furthermore, KNOX gene expression in leaf primordia accompanies leaf dissection in many species, suggesting a role for KNOX genes in making compound leaves (Bharathan et al.,2002).

PHAN is reported to be a negative regulator of KNOXgenes. Mutations in PHAN orthologs (RS2 in maize and AS1 in Arabidopsis) caused KNOX genes to be expressed ectopically (Byrne et al.,2000; Schneeberger et al.,1998; Timmermans et al.,1999; Tsiantis et al.,1999). The phenotype of the double mutant, stm/stm,as1/as1 indicates that as1 is epistatic to stm in Arabidopsis (Byrne et al.,2000). Because AS1 represses KNAT1 (and RS2 represses RS1) and STM in turn represses AS1, the expression domains of PHAN orthologs and KNOX genes do not overlap (Byrne et al., 2000; Timmermans et al., 1999; Tsiantis et al.,1999; Waites et al.,1998). PHAN orthologs are expressed only in the incipient leaf primordium (P₀) and developing leaf primordia(Timmermans et al., 1999; Tsiantis et al., 1999; Waites et al., 1998), but KN1 and STM are expressed in the SAM and are downregulated in P₀ and leaf primordia in species with simple leaves(Jackson et al., 1994; Long et al., 1996). However,in tomato LePHAN and LeT6 mRNA were both detected in the SAM, in leaflet primordia and in growing leaflet laminas(Chen et al., 1997; Janssen et al., 1998; Koltai and Bird, 2000).

We describe four non-allelic mutants, wiry (w), wiry3 (w3), wiry4 (w4) and wiry6(w6) that are defective in ab-adaxial symmetry in tomato. The degree of leaf compounding in these mutant plants was severely reduced. The expression patterns of LeT6, TKN1, LePHAN and LeYAB B were determined in the w, w3 and w6 mutants. The regulatory relationship between LePHAN and KNOX genes in the meristem and early leaf primordium is different from that seen during the later stages of leaf development in tomato and may explain the compound nature of the tomato leaf.

Homozygous mutant seeds of w3, w6(Rick and Butler, 1956), Me and Cu seeds were obtained from the Tomato Genetics Resource Center (TGRC) at the University of California at Davis (accession number: w3 LA1498 cv. First Early, w6 LA2065 cv. Rheinland Rhum, Me/Me LA0324 cv. Rutgers and Cu/Cu LA0325 cv. Stocksdale). For w (Rick and Butler, 1956) and w4(Clayberg et al., 1966),F₂ seeds of self-pollinated heterozygotes (accession number: w/+ LA0274 cv. Canary Export, and w4/+ LA2-237 cv. Pearson)were obtained from TGRC and among the F₂ plants, w/w and w4/w4 plants were examined. Tomato cotyledons of cv. VF36 were transformed according to published protocols(McCormick, 1991) to generate 35SPHAN antisense plants (Kim et al.,2003). All plants were grown in a growth chamber at 22°C with 65% relative humidity and a day length of 16 hours.

Mapping the LePHAN locus and w6

The w and w4 loci are on chromosome four (at 20 cM and 28 cM from the distal end of short arm). The w6 locus was mapped using an F₂ mapping population from a cross between w6 (L. esculentum) and L. pennellii(Tanksley et al., 1989). Using recombination between the w6 mutant phenotype and a LePHANRFLP (HindIII) between L. esculentum and L. pennellii, we determined that the w6 locus is 30 cM from the LePHAN locus on chromosome 10.

Tissues for plastic sections were fixed and sectioned as described previously (Kessler et al.,2001). Samples were viewed with a Nikon Eclipse E600 microscope and images collected using a SPOT (RT Color) digital camera. Samples for SEM were fixed and viewed as described previously(Kessler et al., 2001). Electronic images, collected either directly from the SEM or from a SPOT camera, were processed in Adobe Photoshop.

In situ hybridizations were performed as described previously(Long et al., 1996) using full-length cDNA probes for LeT6, TKN1, LeYAB B and LePHAN. Approximately 500,000 pfu of a λgt10 library from 6-7 mm tomato flowers were screened using INNER NO OUTER, a YABBY member, as probe(Villanueva et al., 1999) to obtain LeYAB B. Median sections (containing the SAM) from multiple different tissue samples including positive controls were placed on each slide and processed. Each experiment was repeated at least four times. Tissues for RT-PCR in situ hybridizations were embedded, sectioned with a Zeiss Microtome HM340E, and processed as previously described(Long et al., 1996). Instead of an overnight hybridization step, RT-PCR was performed on sections as previously described (Ruiz-Medrano et al.,1999). Primers used for the RT-PCR in situ experiments were designed based on the cDNA sequence of LePHAN and LeT6 as follows:

LePHAN1: 5′ACGAGCAGCGTCTTGTTATACAACTAC3′,

LePHAN2: 5′CCCTTCGTCTAAATCCTTGCAGC3′,

LeT65′: 5′TCTTTAACTAACAATAACAATGCAGAAAC3′,

LeT63′: 5′CCAAAGCAGATTCATGAGAAGAATAG3′.

Immunolocalization was performed as described previously(Jackson et al., 1994) using a polyclonal antibody against ROUGHSHEATH2 [a generous gift from Dr Marja Timmermans, for details on antibody preparation see Kim et al.(Kim et al., 2003)].

Wild-type tomato produces unipinnate compound leaves with 7-9 leaflets(Fig. 1A). w, w6 and w3 plants produced mostly cup-shaped or wire-like leaves, but occasionally produced twisted, irregularly shaped flattened leaves with one or two leaflets (Fig. 1B,C,E). In the compound leaves of the w, w3 and w6 mutants, there were 27%, 34% and 19.9% leaflets respectively, compared to wild type (100%; Table 1). The incidence of cup-shaped or wire-like leaves increased in later stages of plant development. A unique morphology was often seen in w3 leaves. These leaves subtended an axillary bud. After production of one or two leaflet pairs, the rachis split and each branch produced an almost complete compound leaf. Often at the junction of the split an axillary-bud like structure was seen(Fig. 1C,D). The w3and w6 mutant plants produced cup-shaped leaves. In contrast, w mutants made tendril-like terminal leaflets. The w, w3 and w6 mutant plants formed normal axillary buds in the axils of the wire-like leaves (Fig. 1F).

To determine if wire-like leaves were produced by abaxialization or adaxialization, the anatomy of these leaves were examined. All parts of a wild-type leaf (including petiole and rachis) have distinct ab-adaxiality. Vascular bundles in the tomato leaf are amphiphloic with both abaxial and adaxial phloem flanking the central xylem(Fig. 2A). Elongated palisade mesophyll cells are located in the adaxial side of the leaf and spongy mesophyll cells are present in the abaxial region of the leaf lamina(Fig. 2B). The w, w3and w6 wire-like leaves were radially symmetric(Fig. 2E-H). This anatomy differed both from the wild-type stem, with a cylinder of vascular tissue surrounding a central pith, and from the wild-type petiole, with clear ab-adaxial symmetry (Fig. 2C,D). Vascular bundles of w, w3 and w6 leaves often had xylem in the center encircled by phloem(Fig. 2E,G,H). Mesophyll cells surrounded the central solid vascular cylinder, but did not have features of distinct elongated palisade mesophyll cells(Fig. 2E-H). In w3leaves producing ectopic leaves with axillary buds, the primary rachis, prior to splitting, had an incompletely closed ring-shaped vascular bundle (arrow),suggesting that this leaf is chimeric with features of both the leaf and the stem (Fig. 2F).

The expanded and flattened leaves of w3 and w6 often showed abaxial patches on the adaxial side of the leaf. In these abaxial patches palisade mesophyll cells were replaced by spongy mesophyll cells(Fig. 2I,J). The w3and w6 leaf had a semicircular vascular bundle with the inner phloem clustered at one end on the adaxial side(Fig. 2L,M) rather than a horseshoe-shaped vascular bundle in the midrib as in the wild-type leaf(Fig. 2A). This suggests a reduced adaxial domain in w3 and w6 leaves. The flattened w leaves had normal mesophyll differentiation in the leaf lamina(Fig. 2K). However, in w, vascular bundles in the midrib were reduced and ectopic palisade cells developed on top of the midrib region(Fig. 2N).

Scanning electron microscopy (SEM) revealed that epidermal cell fates were altered in w3 and w6 leaves. The adaxial epidermal cells of the wild-type leaf were less lobed with fewer crenulations and very few stomata (Fig. 3A), while the abaxial epidermal cells were highly crenulated and irregularly zigzag-shaped with lots of stomata (Fig. 3B). In addition, the wild-type adaxial leaf surface was smooth, compared to the rougher abaxial leaf surface. In contrast, both the upper and lower epidermal cells of w3 leaves had characters intermediate between those seen in the abaxial and adaxial surface of wild type. Both epidermal cells were less lobed (like wild-type adaxial epidermal cells) and had more crenulations (like wild-type abaxial epidermal cells) with roughly equal numbers of stomata,suggesting the loss of distinct abaxial-adaxial epidermal differentiation(Fig. 3C,D). In w6,epidermal cells on both leaf surfaces were highly crenulated and irregular in shape, suggesting abaxialization of the adaxial epidermis of the leaves(Fig. 3E,F). However, w epidermal cells were normal with distinct ab-adaxial features(Fig. 3G,H).

To see if other lateral organs were also abaxialized, we examined floral organs in w, w3 and w6. Tomato flowers have five sepals,petals and stamens and two fused carpels. The bases of sepals are fused into a cup-shaped structure. The corolla is tubular and anthers are adnate to the corolla tube (Fig. 4A). w,w3 and w6 flowers usually had extra floral organs (e.g. 7-10 sepals and petals) and lacked the fusion of floral organs seen in normal flowers. The lateral organs of w, w3 and w6 flowers were narrower than those of wild type (Fig. 4B-D).

In wild-type petals, the adaxial epidermis has protruding papillar cells and no trichomes, while the abaxial epidermis has flattened cells and many trichomes (Fig. 4E,I,J). Vascular bundle organization was not altered in the w6 petal(Fig. 4F). However, papillar cells of the w6 adaxial epidermis had interspersed trichomes. Moreover, the boundary between papillar cells and flattened cells was moved from the margin of the petal toward the adaxial side, indicating a reduction in the adaxial domain of the w6 petal(Fig. 4K). In w3plants, petals were even more abaxialized than in w6 petals; both the abaxial and adaxial epidermal surfaces of w3 petal had flattened cells and numerous trichomes (Fig. 4L,M). The w3 petal had all inner cells packed tightly,resembling the collenchyma cells of the midvein regions(Fig. 4G). Frequently, w petals were radially symmetric and among the papillar cells abaxial patches of flattened cells were seen (Fig. 4H,N). A summary of wiry mutant phenotypes is given in Table 2.

phan mutant plants have a reduced adaxial domain in leaves. We determined if the abaxialized phenotypes and reduced leaflet formation in w, w3 and w6 are due to defects in PHAN expression. Southern blot analysis showed one copy of LePHAN in the tomato genome(data not shown). The chromosome location of LePHAN does not coincide with that of w, w4 and w6. LePHAN mRNA expression was determined by conventional in situ hybridization and by RT-PCR in situ hybridization in the shoot apices and leaves of wild-type, w, w3 and w6 plants.

In the wild-type apex, LePHAN mRNA levels were severalfold higher in the leaf primordia than in the SAM central zone. During early leaf development in wild type, LePHAN transcripts were detected in both adaxial and abaxial sides of the leaf primordium(Fig. 5A), but later, as the leaf primordium grew out, LePHAN mRNA was confined to the adaxial side (Fig. 5B). At later developmental stages, strong LePHAN expression was detected in the leaflet primordium and immature leaflet lamina regions(Fig. 5C).

In the w6 SAM, reduced LePHAN expression was detected in the central zone. Moreover, LePHAN expression in the early leaf primordium, especially on the adaxial side, was much reduced compared to that in wild type (Fig. 5G). No LePHAN was detected in later stages of leaf development of wire-like leaves, but the LePHAN transcript was detected in growing leaflet primordia and leaflet blades in the developing cup-shaped leaves and in leaves with reduced leaflet numbers (Fig. 5G inset). In w and w3 plants, no alteration of LePHAN mRNA accumulation was detected in the SAM and early stage leaf primordia, but LePHAN expression was absent in the later stages of wire-like leaves (Fig. 5L,P and insets).

To further characterize ab-adaxiality in w, w3 and w6leaves, we examined the expression of LeYAB B, a member of the YABBY gene family, in leaves of wild-type and the wiry mutants. LeYAB B was expressed in the abaxial regions of wild-type leaf primordia (Fig. 5D). In w6 plants, LeYAB B expression was seen in both the adaxial and abaxial sides of later radial leaves, and serial sections showed a hollow tube-like pattern of expression (Fig. 5H,I arrow). Earlier flattened w6 leaves showed a wild-type LeYAB B expression pattern(Fig. 5H,I arrowheads). Interestingly, in the w3 leaf primordium, LeYAB B mRNA accumulated on the adaxial instead of the abaxial side(Fig. 5M arrow). No LeYAB B expression was detected in the w leaf primordium(Fig. 5Q). These results were confirmed by northern hybridization to RNA extracted from wiry shoot apices (data not shown).

To determine if reduced leaflet formation in w3 and w6leaves is due to the alteration of class I KNOX gene expression and to determine the regulatory relationship between LePHAN and KNOX genes in tomato, the mRNA expression patterns of two class I KNOX genes, LeT6 (the tomato STM ortholog) and TKN1 (the tomato KNAT1 ortholog) were examined.

In wild type, LeT6 mRNA accumulates in the SAM, in the early leaf primordia, and later in leaflet primordia and growing leaflet blades(Fig. 5E)(Chen et al., 1997). Strong LeT6 expression was detected in the central zone of wild-type SAM(Fig. 5E). In the wand w3 mutants less LeT6 mRNA was detected in the region of the SAM (Fig. 5N,R). Downregulation of LeT6 mRNA was seen in later stages of w,w3 and w6 leaf development. This is equivalent to the stage producing leaflet primordia and growing leaflet lamina in wild type. No LeT6 mRNA could be seen in w6 plants that were producing wire-like leaves (Fig. 5Jinset, R inset). However, LeT6 mRNA localized in the leaflet and leaflet lamina regions of w, w3 and w6 plants that were producing leaves that either were cup-shaped or had a reduced number of leaflets (Fig. 5N insets).

TKN1 expression could be seen in the wild-type SAM, leaf primordia and growing leaflet lamina, but the signal was stronger in the leaf primordia and the peripheral zone of the SAM than in the central zone(Fig. 5F). In w6plants, a high level of TKN1 RNA was detected throughout the SAM and in both early and late leaf primordia, including the radially symmetrical primordia (Fig. 5K). In w3 and w, TKN1 expression was normal in the SAMs but downregulated in the leaflet primordia(Fig. 5O,S). Expression of TKN1 was absent at the tip of the leaf primordium (distal region),where abaxialized wire-like structures are seen in w shoots(Fig. 5S arrow).

The expressions of KNOX genes (LeT6 and TKN1)were altered in wiry mutants. In particular, downregulation of LeT6 in later stage of leaf primordia was accompanied by reduction of leaf compounding in wiry mutants. Reduction of LePHANexpression and upregulation of TKN1 in w6 suggests a negative regulatory relationship between LePHAN and TKN1.

To determine if LeT6 regulates LePHAN in tomato, we analyzed LePHAN expression in Curl (Cu), a mutant known to overexpress LeT6 (Parnis et al., 1997). As reported(Parnis et al., 1997), ectopic expression of the LeT6 mRNA was detected in Cu leaflets and leaflet lamina (Fig. 6A). In Cu plants, LePHAN was present but reduced in the leaf primordia, leaflet and leaflet blade regions(Fig. 6B,C). This LePHAN downregulation in Cu was not sufficient to cause a LePHAN downregulation phenotype, as the Cu leaf showed normal anatomy and epidermal cells (data not shown). Another LeT6overexpression mutation, Mouse Ears (Me), is caused by a gene duplication that leads to early overexpression of a homeobox-containing fusion RNA (Chen et al., 1997; Janssen et al., 1998; Parnis et al., 1997). In the Me mutant, LePHAN expression was reduced and the location of expression was altered (Fig. 6E,F). In the Me plants, LePHAN expression was reduced in the proximal region of the leaf primordia (data not shown) and confined to a narrower adaxial domain in leaf primordia(Fig. 6E). Often, LePHAN expression was absent from the leaf primordia of Me/Me(Fig. 6F), except in vascular tissues, and the leaves produced were radial. This downregulation of LePHAN correlated with the production simple leaves and wire-like leaves at the upper nodes in Me/Me plants(Fig. 6O), phenocopying LePHAN downregulation phenotypes(Fig. 6J). In these LePHAN antisense transgenic plants, LePHAN expression was reduced to a narrow domain or only to vascular tissues(Kim et al., 2003), similar to LePHAN expression in Me/Me(Fig. 6F). Together, these data suggest that LeT6 is a negative regulator of LePHAN.

To determine if LeT6 expression is in turn regulated by LePHAN, we made use of a LePHAN antisense transgenic line. Several independent LePHAN antisense transgenic lines showed cup shaped or wire-like leaves (Fig. 6J) and immunolocalization and in situ RT-PCR experiments showed that LePHAN levels were reduced in these plants and petioles were radial (Kim et al., 2003).

When Cu was crossed into a LePHAN antisense transgenic line, the Cu phenotype was less severe, having less curled leaves and often cup-shaped leaves with simple leaf blades(Fig. 6H). The curled leaf phenotypes were confined to distal region of the leaf. These plants showed elongated and radially symmetric petioles(Fig. 6G). These results suggest that LePHAN downregulation phenotype is epistatic to Cu and that the LeT6 overexpression phenotypes of Cu require LePHAN activity. In Me/Me, LeT6overexpression also led to the production of ectopic shoots on the leaves(Fig. 6K,L, asterisk). These ectopic shoots were formed only in the narrow adaxial domains, where LePHAN was expressed (Fig. 6E). Often this adaxial domain converged to a point(Fig. 6L) and ectopic shoots emerged at this point, suggesting that LePHAN activity is required for LeT6 overexpression phenotypes in Me.

Because a phenocopy of PHAN downregulation is seen only in homozygous Me, but not in heterozygous Me plants, the suppression of LePHAN by LeT6 seems to be dosage sensitive. Me/+ plants show a typical KNOX overexpression phenotype with an increase in leaf compounding (Chen et al., 1997; Parnis et al.,1997) (Fig. 6N). Similarly, w6/w6 homozygous plants (with reduced LePHANlevels) generated wire-like leaves (Fig. 6P) while, w6/+ heterozygous plants produced lobed leaves, a phenotype also seen in the plants overexpressing KNAT1 in Arabidopsis (Fig. 6P).

In w, w3 and w6 mutant plants, partial or complete abaxialization of the lateral organs was observed throughout development,suggesting that W, W3 and W6 play important roles in establishing adaxial cell identity in all lateral organs. The tomato leaf primordium develops basipetally (Dengler,1984) and preferential distal abaxialization in the wleaves indicates that W acts during early leaf development (in the distal region), whereas W3 and W6 function later (in the proximal region) in leaf development. Abaxialization of the wire-like leaf in w, w3 and w6 is different from proximalization of the distal region (changing blade domain into sheath domain) seen in maize mutants(Becraft, 1994; Sinha and Hake,1994; Tsiantis et al.,1999) as unlike petioles, wiry leaves are radially symmetrical. However, the occasional production of stem-like leaves that have ectopic leaves with ectopic axillary buds (in w3) suggests that proximalization into stem-like identity can occur in addition to abaxialization. Adaxial cells are thought to be necessary for the induction of axillary buds in Arabidopsis(McConnell and Barton, 1998; McConnell et al., 2001). The abaxialized wire-like leaves of w, w3 and w6 formed normal axillary buds on the adaxial side of the leaf base(Fig. 1F). Sessile(Arabidopsis) and petiolated (tomato) leaves may have different potentials to form axillary buds in their leaf bases, and adaxial cell fate may not be an absolute requirement for axillary bud formation in tomato. This could also account for the presence of ectopic axillary buds in w3mutant leaves.

It has been proposed that the boundary between abaxial and adaxial cell fates is important for lateral lamina outgrowth(Bowman et al., 2002; Lynn et al., 1999; McConnell and Barton, 1998; Timmermans et al., 1998). Reduced adaxial domain is accompanied by significantly reduced leaflet numbers in w, w3 and w6 (Table 1). One explanation for the fewer leaflets in w, w3 and w6 compound leaves is that leaflet primordium formation, like lamina outgrowth, also requires a proper ab-adaxial boundary.

Two aspects of LePHAN expression in tomato set it apart from orthologs in other species. No other PHAN ortholog has been reported to be expressed in the SAM or specifically in the adaxial domain of leaf primordia. At later stages of leaf development, LePHAN is expressed only in the region of leaflet primordium initiation(Fig. 5C), suggesting that LePHAN (like LeT6) might be involved in leaflet formation,or in establishing ab-adaxiality of leaflets. The possible function of LePHAN in leaflet development is also supported by the fact that LePHAN is not expressed in wire-like leaves and localizes to the growing leaflet primordium or leaflet lamina region in cup-shaped, or less compound leaves of w, w3 and w6. Downregulation of LePHAN was seen in the leaf primordium and leaflet primordium in these mutants, suggesting that W, W3 and W6 are positive regulators of LePHAN expression in leaves. In addition, W6may also regulate LePHAN expression positively in the meristem.

Tomato LePHAN expression was reported to be absent from the SAM in one study (Pien et al., 2001)but was seen in the SAM and leaf primordia in a domain that overlaps the KNOX expression domain by others(Koltai and Bird, 2000). Our results indicate that the latter is the case and that LePHAN(Fig. 5A, Fig. 6F,G) and TKN1are expressed most strongly in the peripheral zone of the meristem, whereas LeT6 expresses strongly in the central zone of the meristem(Fig. 5A,E,F). In Arabidopsis, STM is a negative regulator of AS1. This regulatory relationship is conserved to a large extent in tomato. In Cu and Me (LeT6 overexpression mutants), LePHAN was reduced, suggesting that LeT6 is a negative regulator of LePHAN. TKN1 was upregulated in w6 where LePHAN was downregulated. A simple interpretation for the upregulation of TKN1 in w6 is that LePHAN is a negative regulator of TKN1. However, it is unclear how LePHAN and TKN1 express in an overlapping manner in both the SAM and early leaf primordia. Perhaps LePHAN and another gene (gene A) have a mutually exclusive relationship and gene A in turn inhibits TKN1 expression.

The regulatory dynamics between LePHAN, TKN1 and LeT6 in later leaf and leaflet primordia is different from that in the meristem and early leaf primordium. LePHAN, TKN1 and LeT6 all express in the leaflet primordium and all of them are downregulated in the wire-like leaves of w and w3. These expression data imply that the negative regulation of LeT6 on LePHAN seen in the meristem region does not hold in the wild-type leaflet primordium. Rather, LePHAN functions with LeT6 in a coordinate manner. Cu phenotypes were reduced in antiLePHAN/+ plants and Cu and Me phenotypes were confined to the region where LePHAN was expressed, suggesting that the LeT6overexpression phenotype requires LePHAN function. Similarly,downregulation of LePHAN masked TKN1 overexpression phenotypes in w6/w6 and suggests that TKN1 also requires sufficient LePHAN activity in the leaflet primordium in tomato.

A reduced blade phenotype can be seen only in homozygous Me/Meplants (Fig. 6O) and not in heterozygous Me/+ plants (Fig. 6N), implying that a high dose of LeT6 is needed to downregulate LePHAN in tomato. This hypothesis is also supported by the fact that the expression domains where LePHAN and LeT6express strongly do not overlap (Fig. 5A,E). We suggest that low levels of overexpression of either LeT6 or TKN1 in leaf primordia can cause KNOXoverexpression phenotypes (such as increased dissection of leaves, or more lobed or heart shaped leaves with palmate venation), but high levels of LeT6 overexpression might lead to severe LePHANdownregulation, causing a LePHAN downregulation phenotype. Thus, w6/+ heterozygous plants produced highly lobed leaves(Fig. 6P), a phenotype generally attributed to KNOX gene overexpression, whereas w6/w6 homozygous plants generated mostly cup-shaped or wire-like leaves, which is a LePHAN downregulation phenotype(Fig. 6P). Furthermore, this LePHAN downregulation phenotype masks the KNOXoverexpression phenotypes in tomato, because a certain level of LePHAN is required for the KNOX overexpression phenotypes(as seen in Cu crossed to antiLePHAN and Me/Me). This idea is supported by some of the phenotypes seen in tomato plants that overexpress 35S::LeT6. Some 35S::LeT6 transgenic lines showed wire-like radially symmetrical leaves, resembling PHAN downregulation phenotypes, instead of the typical LeT6 overexpression phenotypes with more leaflets. LeT6 overexpression was at much higher level in these plants producing wire-like leaves, than in plants showing leaflet overproliferation phenotypes (Janssen et al.,1998).

We propose a model (Fig. 7)that summarizes how LeT6 and LePHAN are regulated in tomato. Our results suggest that LeT6 and LePHAN have a mutually antagonistic expression pattern and that each is affected by the quantity of the other. Thus, high levels of LePHAN repress LeT6 and similarly high levels of LeT6 repress LePHAN. Our data does not support increase in LeT6 expression by low levels of LePHAN and vice versa. At intermediate levels both these genes express. Since LeT6 is thought to be necessary for meristem formation in higher plants (although this has not been directly demonstrated in tomato),loss of LeT6 gene function or downregulation of LeT6 could be lethal for plants. Low transformation and plant regeneration success in experiments using 35S::LePHAN constructs support this hypothesis (our unpublished data). LePHAN and LeT6 levels are well balanced in the wild-type leaf, producing 7-9 leaflets with normal ab-adaxiality. Weak LeT6 overexpression and LePHAN downregulation lead to LeT6 overexpression phenotypes seen in the 35S::LeT6 plants(Janssen et al., 1998), Me/+ and Cu leaves. We suggest that the as1mutation showing only KNOX overexpression phenotypes in Arabidopsis and the rs2 phenotype in maize can be categorized in this group. Perhaps, in these instances, KNOXoverexpression does not reach a level that would cause leaf lobing or the PHAN downregulation phenotype. Strong KNOX overexpression and LePHAN downregulation cause LePHAN downregulation phenotypes including cup-shaped or wire-like leaves, as seen in the as1 strong allele (Sun et al.,2002), severe 35S::LeT6, Me/Me, w6/w6 and Cu/Cu;antiLePHAN/+ leaves. However, it should be emphasized that a direct interaction between the KNOX genes and PHAN has not been proved and this interaction may involve multiple regulatory steps.

Our results show that, as seen for the Arabidopsis YAB3 gene, LeYAB B is a good marker for abaxial cell fates(Fig. 5D). LeYAB BmRNA was detected throughout w6 leaf primordia(Fig. 6H, I), while LePHAN mRNA was downregulated in the adaxial region of the leaf(Fig. 5G). This suggests that adaxial cells of leaf primordia in w6 are converted into abaxial cells. These results are consistent with the complete abaxialization of adaxial cells of the w6 leaf (Fig. 2G). However, LeYAB B was unable to downregulate LePHAN in the adaxial region of w3 leaf primordium(Fig. 5L,M) and the absence of LeYAB B did not cause ectopic expression of LePHAN in w (Fig. 5P,Q). In w3, LeYAB B was expressed in the adaxial region of the lateral organs(Fig. 5M). However, in w3, the absence of LeYAB B in the abaxial region still allowed cells to have abaxial fate. While, presence of LeYAB B in the adaxial domain did not cause complete abaxialization, the adaxial epidermis attained some abaxial features (Fig. 3C), suggesting that LeYAB B may play a role in the acquisition of abaxial cell fates. yab and fil mutants have been reported to upregulate KNOX gene expression and result in ectopic shoots in Arabidopsis(Kumaran et al., 2002). By contrast in tomato, ectopic expression of LeYAB B in the adaxial region and absence in the abaxial region of the w3 leaf accompanies ectopic bud formation in these leaves. The fact that ectopic expression of LeYAB B in the adaxial region was detected in both w3 and w6 leaf primordia, but LePHAN was downregulated only in w6 leaf primordium, and complete abaxialization of adaxial cells was seen only in w6 leaf all suggest that LePHAN and other adaxial specific genes play a major role in controlling ab-adaxiality, while the YABBY genes might be involved in a downstream part of the cell fate acquisition pathway. Our results suggest that KNOX gene expression is regulated by presence or absence of LePHAN and not LeYAB B in tomato.

The origins and homologies of compound leaves have been a matter of debate. One view is that dicot compound leaves are a homeotic reiteration of simple leaves along the rachis region of a compound leaf(Lacroix and Sattler, 1994; Rutishauser, 1995). In contrast, others proposed that a compound leaf is formed by dissecting or carving a simple leaf, perhaps by inhibition of blade formation in the rachis area (Hagemann, 1984; Kaplan, 1975).

The adaxial domain is necessary for leaflet primordium formation in tomato. This is reminiscent of the situation where the adaxial domain of a leaf primordium is required for normal SAM activity, and is suggestive of some similarity between compound leaves and shoot systems. This similarity is further supported by the presence of KNOX gene expression in the leaflet primordia in all compound-leafed species from ferns and cycads to higher plants (Bharathan et al.,2002).

If the expression of KNOX genes is crucial to make compound leaves, introducing the expression of KNOX genes into the leaf would have been an important evolutionary innovation that led to the occurrence of compound leaves. In Arabidopsis, Antirrhinum and maize (simple leaves), no KNOX gene expression can be seen in the leaf primordia at any stage of leaf development. Perhaps this is due to the fact that PHAN and KNOX have a very tight mutually exclusive regulatory relationship in Arabidopsis, Antirrhinum and maize(Byrne et al., 2000; Schneeberger et al., 1998; Timmermans et al., 1999; Tsiantis et al., 1999; Waites et al., 1998). Our study shows that both KNOX (LeT6 and TKN1) and LePHAN are expressed in leaflet primordia, suggesting that KNOX genes and LePHAN are not mutually exclusive in the tomato leaflet primordium and that their functions might be dependent on each other. Acquisition of a positive regulatory relationship between KNOXgenes and LePHAN in the leaf primordium might be an evolutionarily significant change to introduce leaflet formation in the ancestral simple leaf primordium. In fact, the discovery that the regulation between KNOXgenes and LePHAN of tomato differs from that of simple-leafed species raises several questions. It will be interesting to determine if the positive regulatory relationship between KNOX genes and PHAN is conserved among compound-leafed species and if this positive regulation is responsible for allowing KNOX expression in leaf primordia of compound-leafed species.

The authors thank Drs John Harada, Connie Champagne and Sharon Kessler for critical comments; members of the Sinha lab for advice and help; members of the Rost Lab for help with histology; Dr Nancy Dengler for many helpful discussions; the Charles M. Rick Tomato Genetics Resource Center for seed stocks; and Dr Marja Timmermans for the anti-RS2 antibody. This paper is dedicated to the late Dr Charles M. Rick for inspiring us to study the tomato leaf. This work was supported by NSF IBN-9983063 and IBN-0316877 to N.S.;Elsie Stocking Memorial, Rosalinde Russell, and UC Davis Jastro-Shields Research Fellowships to M.K.; and SHARP (Howard Hughes) Fellowship to T.P.